CN115169466A

CN115169466A - Method and device for drawing image of land, electronic equipment and computer readable medium

Info

Publication number: CN115169466A
Application number: CN202210832106.2A
Authority: CN
Inventors: 郜贺鹏; 韩博洋; 苏义军; 张钧波; 郑宇�; 寄家豪
Original assignee: Jingdong City Beijing Digital Technology Co Ltd
Current assignee: Jingdong City Beijing Digital Technology Co Ltd
Priority date: 2022-07-15
Filing date: 2022-07-15
Publication date: 2022-10-11

Abstract

The application discloses a plot portrayal method, a plot portrayal device, electronic equipment and a computer readable medium, which relate to the technical field of intelligent cities, and the method comprises the following steps: acquiring track data and identification information corresponding to the track data; generating a resident point sequence according to the track data; matching the resident point sequence with preset plot information to generate a plot access sequence; performing feature extraction on the land parcel access sequence based on a preset target dimension to generate space-time features corresponding to each identification information; and generating a plot representation of each plot in the plot access sequence based on each spatiotemporal feature. Based on the obtained plot image, fine-grained management of urban areas can be assisted, support is provided for detecting the transportation of special substances, the abnormity of inter-area circulation, early warning of dangerous areas, accurate advertisement putting of areas, traffic flow analysis and urban planning, the efficiency and the accuracy of scoring and marking of multiple dimensions of the plot are improved, the manpower participation is reduced, and the data processing cost is reduced.

Description

Method and device for rendering land, electronic equipment and computer readable medium

Technical Field

The application relates to the technical field of intelligent cities, in particular to the technical field of urban management, and specifically relates to a method and a device for block image, electronic equipment and a computer readable medium.

Background

People have great difference on the region division mode and granularity of urban space due to different professional backgrounds and business requirements. The method can directly adopt the division modes of administrative regions or a certain range of a certain road section, and the like, but the modes of representation of land parcels formed by different requirements are different. Even if a building is mapped to a parcel, a single category cannot be used to mark parcels. The land parcel may have different functional semantics under different access sequences. For example, for the hazardous chemical industry, six links of production, storage, transportation, use, operation and treatment of hazardous chemicals need to be supervised, all plots related to the hazardous chemicals in a city need to be concerned, and the functions of the related plots are divided, and the same plot may have different functions in different scenes. The accurate division of the functional representation of the land parcel has important significance for the supervision of the dangerous chemical related industries. The city plot function has space-time attributes and complexity, and can have unique and multi-dimensional characteristics in different time periods. The function of the plot is marked in a manual operation mode, and an expert is needed to mark a plurality of dimensions, so that higher cost is needed.

In the process of implementing the present application, the inventors found that at least the following problems exist in the prior art:

the function of the land parcel is marked in a manual operation mode, and an expert is needed to mark a plurality of dimensions of the land parcel, so that the cost is high, and the efficiency and the accuracy are low.

Disclosure of Invention

In view of this, embodiments of the present application provide a method, an apparatus, an electronic device, and a computer-readable medium for plot representation, which can solve the problems that in the prior art, when a manual operation is used to label the functions of a plot, an expert is required to mark multiple dimensions of the plot, and therefore the cost is high, and the efficiency and the accuracy are low.

To achieve the above object, according to an aspect of embodiments of the present application, there is provided a method of imaging a ground block, including:

acquiring track data and identification information corresponding to the track data;

generating a resident point sequence according to the track data;

matching the resident point sequence with preset plot information to further generate a plot access sequence;

performing feature extraction on the land parcel access sequence based on a preset target dimension to generate space-time features corresponding to each identification information;

and generating a plot representation of each plot in the plot access sequence based on each spatio-temporal feature.

Optionally, generating the sequence of dwell points comprises:

determining the parking position and the parking duration of the vehicle according to the track data;

and generating a resident point sequence according to the resident position and the resident duration of the vehicle.

Optionally, before performing feature extraction on the block visit sequence based on a preset target dimension, the method further includes:

dividing the plot access sequence based on natural day or residence time to obtain each sub-plot access sequence; and

the method for extracting the features of the land parcel access sequence based on the preset target dimension to generate the space-time features corresponding to each identification information comprises the following steps:

and respectively extracting the characteristics of the access sequences of the sub-plots based on the time dimension and the space dimension so as to generate the time characteristics and the space characteristics corresponding to the identification information.

Optionally, generating a plot representation of each plot in the plot access sequence based on each spatiotemporal feature comprises:

and determining corresponding identification information for each land parcel in the land parcel access sequence, and calling a pre-training feature extraction module in the deep learning network model to generate a land parcel portrait of each land parcel in the land parcel access sequence based on the space-time feature corresponding to the identification information corresponding to each land parcel in the land parcel access sequence.

Optionally, generating a plot representation of each plot in the plot access sequence comprises:

selecting a plot from each plot access sequence as an anchor point;

screening the plot access sequences passing through the anchor points, and further determining a target plot access sequence from the plot access sequences passing through the anchor points;

and acquiring the space-time characteristics corresponding to the target plot access sequence, and generating a plot portrait of the plot corresponding to the anchor point based on the space-time characteristics corresponding to the target plot access sequence, the anchor point position in the target plot access sequence, the anchor point and the space-time characteristics of the plot access sequence where the anchor point is located.

Optionally, determining a target block access sequence from the block access sequences passing through the anchor point comprises:

determining the length of a land parcel access sequence passing through an anchor point and the position of the anchor point;

based on the length and the anchor position, a target block access sequence is determined from the block access sequences passing through the anchor.

Optionally, before invoking the pre-training feature extraction module in the deep learning network model, the method further includes:

acquiring an initial neural network model;

acquiring a training sample set, wherein the training sample set comprises a similar sample pair based on spatial characteristics, a similar sample pair based on temporal characteristics, a spatial anchor point pair corresponding to the similar sample pair based on the spatial characteristics, a temporal anchor point pair corresponding to the similar sample pair based on the temporal characteristics, a first distance of a labeled spatial anchor point pair, a second distance of the labeled temporal anchor point pair and a labeled plot image;

taking the similar sample pair based on the spatial characteristics as the input of a coding conversion layer of the initial neural network model, taking the spatial anchor point pair corresponding to the similar sample pair based on the spatial characteristics and the first distance as the expected output of the coding conversion layer of the initial neural network model, and carrying out self-supervision training on the initial neural network model;

using the similar sample pair based on the time characteristics as the input of a coding conversion layer of the initial neural network model, using the corresponding time anchor point pair and the second distance of the similar sample pair based on the time characteristics as the expected output of the coding conversion layer of the initial neural network model, and carrying out self-supervision training on the initial neural network model;

and respectively taking the space anchor point pair, the time anchor point pair, the first distance and the second distance as the input of a linear layer of the initial neural network model, taking the expressed image of the land as the expected output of the linear layer, training the initial neural network model, and further optimizing through a loss function to obtain a pre-training feature extraction module in the deep learning network model.

In addition, the present application also provides a plot representation apparatus, comprising:

a receiving unit configured to acquire the trajectory data and identification information corresponding to the trajectory data;

a resident point sequence generating unit configured to generate a resident point sequence according to the trajectory data;

the system comprises a land parcel access sequence generating unit, a land parcel access sequence generating unit and a land parcel access sequence generating unit, wherein the land parcel access sequence generating unit is configured to match a resident point sequence with preset land parcel information so as to generate a land parcel access sequence;

the space-time feature generation unit is configured to perform feature extraction on the land parcel access sequence based on a preset target dimension so as to generate space-time features corresponding to each identification information;

a plot representation generation unit configured to generate a plot representation of each plot in the plot access sequence based on each spatiotemporal feature.

Optionally, the dwell point sequence generation unit is further configured to:

Optionally, the apparatus further comprises a sequence dividing unit configured to:

the spatiotemporal feature generation unit is further configured to:

Optionally, the parcel representation generating unit is further configured to:

Optionally, the parcel representation generation unit is further configured to:

selecting a plot from each plot access sequence as an anchor point;

and acquiring the space-time characteristics corresponding to the target block access sequence, and generating a block portrait of the block corresponding to the anchor point based on the space-time characteristics corresponding to the target block access sequence, the anchor point position in the target block access sequence, the anchor point and the space-time characteristics of the block access sequence where the anchor point is located.

Optionally, the parcel representation generation unit is further configured to:

Optionally, the plot representation apparatus further comprises a model training unit configured to:

acquiring an initial neural network model;

acquiring a training sample set, wherein the training sample set is configured to be a similar sample pair based on spatial features, a similar sample pair based on temporal features, a spatial anchor point pair corresponding to the similar sample pair based on the spatial features, a temporal anchor point pair corresponding to the similar sample pair based on the temporal features, a first distance of a labeled spatial anchor point pair, a second distance of the labeled temporal anchor point pair and a labeled plot image;

In addition, the present application also provides a data processing electronic device, including: one or more processors; a storage device for storing one or more programs which, when executed by one or more processors, cause the one or more processors to implement a method of tile imaging as described above.

In addition, the present application also provides a computer readable medium, on which a computer program is stored, which when executed by a processor, implements the method of mapping a parcel as described above.

One embodiment of the above invention has the following advantages or benefits: the method comprises the steps of obtaining track data and identification information corresponding to the track data; generating a resident point sequence according to the track data; matching the resident point sequence with preset plot information to further generate a plot access sequence; performing feature extraction on the land parcel access sequence based on a preset target dimension to generate space-time features corresponding to each identification information; and generating a plot representation of each plot in the plot access sequence based on each spatiotemporal feature. Performing feature extraction on the land parcel access sequence based on a preset target dimension to generate space-time features corresponding to each identification information; and generating a plot representation of each plot in the plot access sequence based on each spatio-temporal feature. Based on the obtained plot image, fine-grained management of urban areas can be assisted, and support is provided for detecting the transportation of special substances, the abnormity of inter-area circulation, early warning of dangerous areas, accurate advertisement putting of areas, traffic flow analysis and urban planning, so that the efficiency and the accuracy of scoring and marking of multiple dimensions of the plot are improved, the manpower participation is reduced, and the data processing cost is reduced.

Further effects of the above-mentioned non-conventional alternatives will be described below in connection with the embodiments.

Drawings

The drawings are included to provide a further understanding of the application and are not to be construed as limiting the application. Wherein:

FIG. 1 is a schematic diagram of a main flow of a land parcel representation method according to a first embodiment of the present application;

FIG. 2 is a schematic diagram of a main flow of a land parcel representation method according to a second embodiment of the present application;

FIG. 3 is a diagram illustrating an application scenario of a land parcel rendering method according to a third embodiment of the present application;

FIGS. 4a, 4b, and 4c are schematic diagrams of a time-based track sample pair, a space-based track sample pair, and a mixed track sample pair, respectively, of a parcel rendering method according to an embodiment of the present application;

FIG. 5 is a diagram of a pre-trained model framework for a method of mapping a parcel according to an embodiment of the present application;

FIG. 6 is a schematic diagram of the main units of a land parcel photographing apparatus according to an embodiment of the present application;

FIG. 7 is an exemplary system architecture diagram to which embodiments of the present application may be applied;

fig. 8 is a schematic structural diagram of a computer system suitable for implementing the terminal device or the server according to the embodiment of the present application.

Detailed Description

The following description of the exemplary embodiments of the present application, taken in conjunction with the accompanying drawings, includes various details of the embodiments of the application for the understanding of the same, which are to be considered exemplary only. Accordingly, those of ordinary skill in the art will recognize that various changes and modifications of the embodiments described herein can be made without departing from the scope and spirit of the present application. Also, descriptions of well-known functions and constructions are omitted in the following description for clarity and conciseness. According to the technical scheme, the data acquisition, storage, use, processing and the like meet relevant regulations of national laws and regulations.

Fig. 1 is a schematic diagram illustrating a main flow of a method for representing a parcel image according to a first embodiment of the present application, and as shown in fig. 1, the method for representing a parcel image includes:

step S101, acquiring track data and identification information corresponding to the track data.

In this embodiment, an execution subject (for example, a server) of the parcel rendering method may acquire the trajectory data and the identification information corresponding to the trajectory data by wired connection or wireless connection. The trajectory data may be trajectory data during vehicle driving, and the identification information corresponding to the trajectory data may be license plate information, owner name information, and the like of the vehicle corresponding to the acquired trajectory data, and each piece of identification information may correspond to one vehicle. The data processing request may specifically be a request for acquiring a function of the target parcel. The target land may be a land through which a vehicle travel track passes. For example, as shown by the dots in fig. 4a, 4b, and 4c, the dots in the figures may be the target parcel.

And S102, generating a resident point sequence according to the track data.

The execution subject may extract the identification information, the vehicle dwell position, and the dwell time interval (i.e., the dwell time duration) from the trajectory data. The sequence of dwell points for the vehicle includes identification information, a vehicle dwell position, a dwell time interval. Exemplary, the parking spot sequence may be for a hotel (a vehicle, 10 am 10-10 am 02, park 2 minutes) -B gas station (a vehicle, 10 am 10-10 am 15, park 5 minutes) -C school-D library (a vehicle, 11 am 11-pm 13, park 2 hours.

And step S103, matching the resident point sequence with the preset parcel information to further generate a parcel access sequence.

The preset parcel information may be, for example, a preset community range. The community scope includes the change scope of the community over time and the community space area scope. First, the execution subject needs to match the residence position corresponding to the residence point sequence with the community range. According to the sequence of the residence time, the sequence is converted into a sequence of the vehicles accessing the plots corresponding to each community range, namely a plot accessing sequence, for example, the plot accessing sequence may be: parcel a (a vehicle, 10 am-10 am, park 2 min) -parcel B (a vehicle, 10 am-10 am, park 5 min) -parcel C-parcel D (a vehicle, 11 am, 00-pm 13, park 2 h.

And step S104, performing feature extraction on the land parcel access sequence based on a preset target dimension to generate space-time features corresponding to each identification information.

A preset target dimension, such as a time dimension, a space dimension, etc. The feature extraction is performed on the land parcel access sequence based on a preset target dimension, specifically, the feature extraction may be performed on the land parcel access sequence based on a time dimension and a space dimension respectively, so as to generate a time feature and a space feature corresponding to each identification information, which are referred to as a space-time feature for short. Each piece of identification information can correspond to one vehicle, and if a plurality of vehicles access the same plot simultaneously in the plot access sequence, the number of the identification information is multiple.

Specifically, before feature extraction is performed on the parcel access sequence based on a preset target dimension, the parcel representation method further comprises:

and dividing the plot access sequence based on natural day or residence time to obtain each sub-plot access sequence.

In an example, the access sequence of the parcel is segmented into sequences with indefinite lengths, that is, each access sequence of the sub-parcel is obtained by the segmentation. There are various ways of division. The specific division mode is as follows:

1) Divided by natural day, in a daily 04:00 as demarcation points, every day 04:00 to day 04: the sequence formed during 00 was taken as a sample. Based on time division, the travel demands of people are mostly in the minimum period of days. And dividing by days to obtain a sub-parcel access sequence of a complete travel period.

2) And (4) counting all residence time lengths according to the division of the long residence time points. Setting the thresholds of the dimensions of residence time length, residence frequency and the like. And finding frequent long-time residence points as nodes of the segmentation track. And taking the space as a division basis, combining the travel duration of the crowd and different travel time points, and when the crowd passes through certain specific plots, segmenting the plot access sequence to obtain each sub-plot access sequence.

Each sub-parcel access sequence is treated as a sample by slicing the sequence of access parcels. Finally, time slice division is performed.

In the embodiment of the present application, performing feature extraction on a parcel access sequence based on a preset target dimension to generate a spatiotemporal feature corresponding to each identification information includes:

Specifically, the features corresponding to the same time in each sub-parcel access sequence are extracted to obtain the time features of each identification information corresponding to each sub-parcel access sequence. For example, the time characteristics of each identification information corresponding to each sub-parcel access sequence may include identification information, residence time, access behavior, and other characteristics corresponding to vehicles having similar tracks at similar times; and extracting the characteristics corresponding to the same space in each sub-parcel access sequence to obtain the spatial characteristics corresponding to each identification information corresponding to each sub-parcel access sequence. For example, a plurality of vehicles pass through a restaurant, and the identification information, the residence time, the access behavior and other characteristics of the vehicles passing through the same restaurant are extracted spatial characteristics.

Step S105, generating a plot image of each plot in the plot access sequence based on each spatio-temporal feature.

The execution main body executes portrait construction on each corresponding land parcel based on the extracted space-time characteristics corresponding to each piece of identification information, and takes the characteristics of vehicles, residence time, access behaviors and the like passing by each land parcel at the same time point as the constituent parts of the land parcel portrait of each land parcel; and combining the spatial characteristics to obtain the characteristics of vehicles passing through the same land in a preset time period, the residence time, the access behavior and the like, further serving as the supplement of the land image of each land to further perfect the land image, and finally generating the land image of each land in the land access sequence. The plot image can specifically represent information such as vehicle transportation material types, vehicle circulation conditions, plot danger degree and traffic flow of the plot.

In the embodiment, the track data and the identification information corresponding to the track data are obtained; generating a resident point sequence according to the track data; matching the resident point sequence with preset plot information to further generate a plot access sequence; performing feature extraction on the land parcel access sequence based on a preset target dimension to generate space-time features corresponding to each identification information; and generating a plot representation of each plot in the plot access sequence based on each spatio-temporal feature. Performing feature extraction on the land parcel access sequence based on a preset target dimension to generate space-time features corresponding to each identification information; and generating a plot representation of each plot in the plot access sequence based on each spatio-temporal feature. Based on the obtained plot image, fine-grained management of urban areas can be assisted, and support is provided for detecting the transportation of special substances, the abnormity of inter-area circulation, early warning of dangerous areas, accurate advertisement putting of areas, traffic flow analysis and urban planning, so that the efficiency and the accuracy of scoring and marking of multiple dimensions of the plot are improved, the manpower participation is reduced, and the data processing cost is reduced.

Fig. 2 is a schematic main flow chart of a land parcel imaging method according to a second embodiment of the present application, and as shown in fig. 2, the land parcel imaging method includes:

step S201, acquiring the track data and the identification information corresponding to the track data.

And step S202, determining the parking position and the parking duration of the vehicle according to the track data.

And step S203, generating a resident point sequence according to the resident position and the resident duration of the vehicle.

Specifically, the execution subject may package the vehicle parking position and the parking duration information in the parking point, package each parking position of the vehicle and the corresponding parking duration in one parking point, and further generate the parking point sequence according to the time sequence.

And step S204, matching the resident point sequence with preset plot information to further generate a plot access sequence.

And S205, performing feature extraction on the land parcel access sequence based on a preset target dimension to generate space-time features corresponding to each identification information.

And S206, determining corresponding identification information for each land parcel in the land parcel access sequence, and calling a pre-training feature extraction module in the deep learning network model to generate a land parcel portrait of each land parcel in the land parcel access sequence based on the spatiotemporal feature corresponding to the identification information corresponding to each land parcel in the land parcel access sequence.

Specifically, generating a plot representation of each plot in a plot access sequence comprises:

selecting a land parcel as an anchor point from each land parcel access sequence; for each land access sequence, a land can be selected from a land access sequence as an anchor point. The selected anchor point may then characterize this entire sequence of block accesses. The position of an anchor point is the position of a certain block in a block access sequence.

Then, executing a main body to screen the land parcel access sequences passing through the anchor points, and further determining a target land parcel access sequence from the land parcel access sequences passing through the anchor points; the execution main body can screen the plot access sequence passing through the anchor point after selecting the anchor point, so that which plot access sequences have intersection with the plot access sequence where the anchor point is located can be determined, the possibility that the plot access sequence where the anchor point is located and which plot access sequences have similarity can be determined, and the plot image of the plot corresponding to the anchor point can be enriched. The target parcel access sequence can be a parcel access sequence in which the quantity of parcels in the parcel access sequence passing through the anchor point exceeds a threshold value, that is, the parcel access sequence can be used as the target parcel access sequence after the quantity of parcels in the parcel access sequence passing through the anchor point reaches a certain quantity so as to enrich the parcel portraits of the parcels corresponding to the anchor point.

Specifically, determining a target block access sequence from the block access sequences passing through the anchor point comprises:

The execution body may first determine a length L1 and a position P1 of a land mass access sequence where the anchor O is located, then the execution body may determine a length of the land mass access sequence passing through the anchor O and an anchor position, and then the execution body may determine a land mass access sequence having a length (e.g., L2, L3, L4 \8230;) of the land mass access sequences passing through the anchor O and a length L1 and a position P1 of the anchor position (P2, P3, P4 \8230;) similar to the land mass access sequence where the anchor O is located as a target land mass access sequence. That is, the lengths and anchor positions of the target block access sequences (e.g., S1, S2) are similar to the length L1 and anchor position P1 of the block access sequence at which the current anchor O is located, where "similar" means that the difference between the sequence lengths is less than a preset length threshold and the difference between the anchor positions is less than a preset distance threshold.

Specifically, before calling a pre-training feature extraction module in the deep learning network model, the method for mapping the land parcel further comprises the following steps:

the initial neural network model is obtained, for example, may be a transform model, or may also be a model such as RNN, LSTM, TCN, and the like.

For example, when the training sample set is obtained, the spatio-temporal sample pair may be specifically constructed. Training samples are obtained through data preprocessing, and a contrast learning sample structure mainly constructs contrast samples through two dimensions of time and space based on sequence data.

The space-based construction mode is as follows: and selecting sequences accessed by the plots in which the same plot is located to form a candidate set, and matching the sequences according to the similarity of dimensions such as vehicles, sequence lengths and the like of the generated sequences to generate similar sample pairs. Meanwhile, to limit the similarity of the two samples forming the similar sample pair, if the similarity of the two samples is too high, it is meaningless for training, and the samples need to be filtered. The construction example is shown in fig. 4b based on a spatial construction mode, two sequence E nodes are anchor points and labeled as residence time periods, and the specific steps are as follows:

any field ld (i.e., field id, which may be, for example, a field number) is selected as an anchor point, and the sequence S of all field samples passing through the field ld is screened.

And setting a threshold V according to the length of the sample sequence and the ordinal number of the sequence of the anchor point ld, and when the threshold V is reached, considering that the two samples can form a similar sample pair to form a candidate set Sn.

And respectively forming a block sequence for two samples in the candidate set Sn, and calculating the longest similar subsequence. A threshold W is set and when the similarity is above the threshold W, the candidate set Sn clears this comparison sample pair.

The candidate set Sn is the final constructed similar comparison sample space. As shown in fig. 4b, i.e. a similar contrast sample space is constructed, i.e. a similar sample pair based on spatial features.

The time-based construction mode is as follows: similar sample pair construction based on time characteristics is performed according to the tide nature of vehicle travel. Sequences of visiting places made up of vehicles of the same type (route-defining vehicles other than buses) have similar semantics. There are a number of ways to construct the similarity according to the time dimension. The method mainly comprises the following steps: different sequences of the same vehicle, two plots in the two sequences are accessed in similar time periods, tracks have higher similarity (for example, the dwell time of the accessed plots is similar), and behavior of the two sample sequences accessing the plots has higher similarity. The configuration is as shown in fig. 4a, for example, in fig. 4a, the reference numeral 09. 08. In fig. 4a, two sequence F and G nodes are anchor points.

As another implementation manner of the embodiment of the application, the training sample set may also combine two dimensions of time and space at the same time to construct similar comparison samples. As shown in fig. 4c, the construction example includes the following specific steps that E nodes of the two sequences are anchor points and labeled as residence time periods:

any plot ld is selected as an anchor point, and all plot samples passing through the plot ld are screened for sequence S.

The sequence S of plot samples is partitioned by accessing the time interval of the plot ld. And accessing sequences with similar time intervals and dividing the sequences into candidate sets of similar sequences.

And setting a threshold V according to the length of the sample sequence and the ordinal number of the sequence where the anchor point ld is positioned as characteristics, and considering that the two samples can form a similar sample pair to form a candidate set Sn when the threshold V is reached.

The candidate set Sn is a similar sample pair based on spatiotemporal characteristics which is finally constructed.

The execution main body can input the space-time based similar sample pair to the initial neural network model, and the plot image corresponding to the marked anchor point is used as expected output so as to train the initial neural network model, and finally the pre-training feature extraction module in the deep learning network model is obtained.

It can be understood that the essence of the pre-training feature extraction module in the deep learning network model is to cluster the input block visit sequences based on the temporal features and the spatial features, to shorten the distance of the anchor points of the similar block visit sequences in the model space, and to increase the distance of the anchor points of the dissimilar block visit sequences in the space. Thus, the parcel representation corresponding to the anchor point of the input parcel access sequence is generated by clustering. And the plot image of the plot R is used for representing the information such as identification information, vehicle residence time, vehicle access behavior and the like of the plot R accessed at each time point of the time period R.

After the training sample set is acquired, a model training process may be performed. The method comprises the following specific steps:

when the execution subject constructs samples which can be used for the self-supervision contrast learning, namely similar samples based on the spatial features and similar samples based on the spatial features, the model can be trained. The method and the device for pre-training the sequence model are used for modeling the sequence model, and a transformer frame structure is selected for pre-training the characterization of the land parcel.

The pre-training model framework is shown in FIG. 5: wherein the input part is p _a ¹ ，p _a ² ，……，p _a ^x ，……，p _a ⁿ And p _b ¹ ，p _b ² ，……，p _b ^y ，……，p _b ^m For two input sample data, p represents a plot in the sequence, the upper right corner is labeled as the ordinal number of the plot p in the sample sequence, and the lower right corner is labeled as the sequence number. The lengths of the two sequences a, b are n, m,the anchor points are plots at x and y positions, respectively. The model part is a coding conversion layer (transform encoding layer) and a linear layer (linear layer), and the transform part can stack N layers according to the actual situation of the model to optimize the model. The transformer structures of the left and right parts in FIG. 5 are the same and share parameters. The initial block code passes through N layers of transformers, and the output is a sequence with the same length as the input sequence, and the dimensions of the two sequences a and b are N x d and m x d respectively. And selecting the representations of the anchor points x and y to respectively pass through the linear layer, and outputting the final representation of the land parcel to obtain the image of the land parcel. Finally, optimization is performed through the loss function. Similar samples are drawn closer in distance in model space, and negative samples are increased in distance in space. The loss function may be a sum of squares loss function, a hind loss function, or the like.

After the model is pre-trained, the model can be used for model reasoning, specifically, a linear layer can be removed, and parameters of a pre-training transformer part are frozen. And a downstream network structure is designed by combining the characteristics of downstream tasks, and the downstream network can be finely adjusted by using small samples. After fine tuning, specific downstream tasks can be predicted. A land parcel image pre-training mode based on self-supervision and comparison of a track sequence is constructed, and the problems that land parcel labels are few and large-scale models are difficult to train are solved. The functional characterization and classification of the same land can be obtained under different time and access sequences, and more stable and universal land characterization can be provided for downstream tasks.

FIG. 3 is a diagram illustrating an application scenario of a method for representing a parcel of land according to a third embodiment of the present application. The plot portrayal method is applied to scenes of functional representation and classification of the same plot under different time and access sequences. In the embodiment of the present application, POI: point of interest; AOI: area of interest; OD: the start and end points of the trajectory; loss: a loss function. The city plots are spatial ranges obtained by dividing the cities. The spatial granularity of the plot can be differentiated according to downstream tasks and different requirements. The function of a parcel is mainly reflected in the composition and number of categories of POIs within a parcel, and different POIs of the same category have their unique properties. Classifying and labeling urban plots is complex and relatively difficult to quantify across dimensions. And the OD (starting and ending point) and the dwell point of people and vehicles have extremely high correlation with urban plots. In ecology, the granularity includes spatial granularity and temporal granularity, and the spatial granularity represents a characteristic length, an area or a volume represented by a minimum recognizable unit in a space of a research area, such as a grid pixel of an image of the research area, a sample prescription and the like; temporal granularity refers to the frequency or time interval at which a certain object or event under study occurs (or is sampled).

The embodiment of the application uses a vehicle OD sequence or track (needing to be preprocessed into a resident point sequence) to classify and label the land parcel. The method framework is shown in FIG. 3, and the modeling process comprises data preprocessing, model training and model reasoning. The final output is an embedded representation of the plot, i.e., a plot representation, which may be used to support downstream tasks such as region recommendation, city planning, anomaly detection, traffic flow, etc. Specifically, as shown in fig. 3, trajectory data is first acquired. The trajectory data mainly comprises a vehicle dwell point sequence, a community area range and the like. And preprocessing the track data to obtain a resident point sequence, and inputting the obtained resident point sequence into a pre-training feature extraction module in the deep learning network model to output a plot image. The data preprocessing comprises the steps of matching the resident positions in the input data with communities, converting the resident positions into access sequences of vehicles to all plots according to the resident time sequence, and further segmenting the access sequences of the plots to be divided into sequences with indefinite lengths. The functional characterization and classification of the same land can be obtained under different time and access sequences, and more stable and universal land characterization can be provided for downstream tasks.

FIG. 6 is a schematic diagram of the main units of a land parcel photographing apparatus according to an embodiment of the present application. As shown in fig. 6, the parcel rendering apparatus 600 includes a receiving unit 601, a resident point sequence generating unit 602, a parcel access sequence generating unit 603, a spatio-temporal feature generating unit 604, and a parcel rendering generating unit 605.

The receiving unit 601 is configured to acquire the track data and the identification information corresponding to the track data.

A dwell point sequence generating unit 602 configured to generate a dwell point sequence according to the trajectory data.

A parcel access sequence generating unit 603 configured to match the dwell point sequence with preset parcel information, thereby generating a parcel access sequence.

And a spatiotemporal feature generation unit 604 configured to perform feature extraction on the parcel access sequence based on a preset target dimension to generate spatiotemporal features corresponding to each identification information.

A parcel representation generation unit 605 configured to generate a parcel representation of each parcel in the parcel access sequence based on each spatiotemporal feature.

In some embodiments, the dwell point sequence generation unit 602 is further configured to: determining the parking position and the parking duration of the vehicle according to the track data; and generating a resident point sequence according to the resident position and the resident duration of the vehicle.

In some embodiments, the apparatus further comprises a sequence dividing unit, not shown in fig. 6, configured to: dividing the plot access sequence based on natural day or residence time to obtain each sub-plot access sequence; and the spatio-temporal feature generation unit is further configured to: and respectively extracting the characteristics of the access sequences of the sub-plots based on the time dimension and the space dimension so as to generate the time characteristics and the space characteristics corresponding to the identification information.

In some embodiments, parcel representation generation unit 605 is further configured to: and determining corresponding identification information for each land parcel in the land parcel access sequence, and calling a pre-training feature extraction module in the deep learning network model to generate a land parcel portrait of each land parcel in the land parcel access sequence based on the space-time feature corresponding to the identification information corresponding to each land parcel in the land parcel access sequence.

In some embodiments, parcel representation generation unit 605 is further configured to: selecting a plot from each plot access sequence as an anchor point; screening the land parcel access sequences passing through the anchor points, and further determining a target land parcel access sequence from the land parcel access sequences passing through the anchor points; and acquiring the space-time characteristics corresponding to the target plot access sequence, and generating a plot portrait of the plot corresponding to the anchor point based on the space-time characteristics corresponding to the target plot access sequence, the anchor point position in the target plot access sequence, the anchor point and the space-time characteristics of the plot access sequence where the anchor point is located.

In some embodiments, parcel representation generation unit 605 is further configured to: determining the length of a land parcel access sequence passing through an anchor point and the position of the anchor point; based on the length and the anchor position, a target block access sequence is determined from the block access sequences passing through the anchor.

In some embodiments, the parcel representation apparatus further comprises a model training unit, not shown in FIG. 6, configured to: acquiring an initial neural network model; acquiring a training sample set, wherein the training sample set is configured to be a similar sample pair based on spatial features, a similar sample pair based on temporal features, a spatial anchor point pair corresponding to the similar sample pair based on the spatial features, a temporal anchor point pair corresponding to the similar sample pair based on the temporal features, a first distance of a labeled spatial anchor point pair, a second distance of the labeled temporal anchor point pair and a labeled plot image; taking the similar sample pair based on the spatial characteristics as the input of a coding conversion layer of the initial neural network model, taking the spatial anchor point pair corresponding to the similar sample pair based on the spatial characteristics and the first distance as the expected output of the coding conversion layer of the initial neural network model, and carrying out self-supervision training on the initial neural network model; using the similar sample pair based on the time characteristics as the input of a coding conversion layer of the initial neural network model, using the corresponding time anchor point pair and the second distance of the similar sample pair based on the time characteristics as the expected output of the coding conversion layer of the initial neural network model, and carrying out self-supervision training on the initial neural network model; and respectively taking the space anchor point pair, the time anchor point pair, the first distance and the second distance as the input of a linear layer of the initial neural network model, taking the expressed image of the land as the expected output of the linear layer, training the initial neural network model, and further optimizing through a loss function to obtain a pre-training feature extraction module in the deep learning network model.

In the present application, the parcel image method and parcel image apparatus have a corresponding relationship in the details of implementation, and therefore, the description of the details thereof will not be repeated.

Fig. 7 illustrates an exemplary system architecture 700 to which the method or apparatus of land parcel rendering of embodiments of the present application may be applied.

As shown in fig. 7, the system architecture 700 may include

terminal devices

701, 702, 703, a network 704, and a server 705. The network 704 serves to provide a medium for communication links between the

terminal devices

701, 702, 703 and the server 705. Network 704 may include various connection types, such as wired, wireless communication links, or fiber optic cables, to name a few.

A user may use the

terminal devices

701, 702, 703 to interact with a server 705 over a network 704, to receive or send messages or the like. The

terminal devices

701, 702, 703 may have installed thereon various communication client applications, such as a shopping-like application, a web browser application, a search-like application, an instant messaging tool, a mailbox client, social platform software, etc. (by way of example only).

The

terminal devices

701, 702, 703 may be various electronic devices having a data processing screen and supporting web browsing, including but not limited to smart phones, tablet computers, laptop portable computers, desktop computers, and the like.

The server 705 may be a server providing various services, such as a background management server (for example only) providing support for data processing requests submitted by users using the

terminal devices

701, 702, 703. The background management server can acquire the track data and identification information corresponding to the track data; generating a resident point sequence according to the track data; matching the resident point sequence with preset plot information to further generate a plot access sequence; performing feature extraction on the land parcel access sequence based on a preset target dimension to generate space-time features corresponding to each identification information; and generating a plot representation of each plot in the plot access sequence based on each spatio-temporal feature. Performing feature extraction on the land parcel access sequence based on a preset target dimension to generate space-time features corresponding to each identification information; and generating a plot representation of each plot in the plot access sequence based on each spatiotemporal feature. Based on the obtained plot image, fine-grained management of urban areas can be assisted, and support is provided for detecting the transportation of special substances, the abnormity of inter-area circulation, early warning of dangerous areas, accurate advertisement putting of areas, traffic flow analysis and urban planning, so that the efficiency and the accuracy of scoring and marking of multiple dimensions of the plot are improved, the manpower participation is reduced, and the data processing cost is reduced.

It should be noted that the method for representing a parcel of land provided by the embodiment of the present application is generally executed by the server 705, and accordingly, a parcel representing apparatus is generally installed in the server 705.

It should be understood that the number of terminal devices, networks, and servers in fig. 7 is merely illustrative. There may be any number of terminal devices, networks, and servers, as desired for an implementation.

Referring now to FIG. 8, shown is a block diagram of a computer system 800 suitable for use in implementing a terminal device of an embodiment of the present application. The terminal device shown in fig. 8 is only an example, and should not bring any limitation to the functions and the scope of use of the embodiments of the present application.

As shown in fig. 8, a computer system 800 includes a Central Processing Unit (CPU) 801 which can perform various appropriate actions and processes in accordance with a program stored in a Read Only Memory (ROM) 802 or a program loaded from a storage section 808 into a Random Access Memory (RAM) 803. In the RAM803, various programs and data necessary for the operation of the computer system 800 are also stored. The CPU801, ROM802, and RAM803 are connected to each other via a bus 804. An input/output (I/O) interface 805 is also connected to bus 804.

The following components are connected to the I/O interface 805: an input portion 806 including a keyboard, a mouse, and the like; an output section 807 including a display such as a Cathode Ray Tube (CRT), a liquid crystal credit authorization query processor (LCD), and the like, and a speaker; a storage section 808 including a hard disk and the like; and a communication section 809 including a network interface card such as a LAN card, a modem, or the like. The communication section 809 performs communication processing via a network such as the internet. A drive 810 is also connected to the I/O interface 805 as needed. A removable medium 811 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, or the like is mounted on the drive 810 as necessary, so that a computer program read out therefrom is mounted on the storage section 808 as necessary.

In particular, according to embodiments disclosed herein, the processes described above with reference to the flow diagrams may be implemented as computer software programs. For example, embodiments disclosed herein include a computer program product comprising a computer program embodied on a computer readable medium, the computer program comprising program code for performing the method illustrated by the flow chart. In such an embodiment, the computer program can be downloaded and installed from a network through the communication section 809 and/or installed from the removable medium 811. The computer program executes the above-described functions defined in the system of the present application when executed by the Central Processing Unit (CPU) 801.

It should be noted that the computer readable medium shown in the present application may be a computer readable signal medium or a computer readable storage medium or any combination of the two. A computer readable storage medium may include, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the foregoing. More specific examples of the computer readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the present application, a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. In this application, however, a computer readable signal medium may include a propagated data signal with computer readable program code embodied therein, for example, in baseband or as part of a carrier wave. Such a propagated data signal may take any of a variety of forms, including, but not limited to, electro-magnetic, optical, or any suitable combination thereof. A computer readable signal medium may be any computer readable medium that is not a computer readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device. Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, fiber optic cable, RF, etc., or any suitable combination of the foregoing.

The flowchart and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present application. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams or flowchart illustration, and combinations of blocks in the block diagrams or flowchart illustration, can be implemented by special purpose hardware-based systems that perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.

The units described in the embodiments of the present application may be implemented by software or hardware. The described units may also be provided in a processor, and may be described as: a processor includes a receiving unit, a dwell point sequence generating unit, a parcel access sequence generating unit, a spatiotemporal feature generating unit, and a parcel sketch generating unit. Wherein the names of the elements do not in some way constitute a limitation on the elements themselves.

As another aspect, the present application also provides a computer-readable medium, which may be contained in the apparatus described in the above embodiments; or may be separate and not incorporated into the device. The computer readable medium carries one or more programs, and when the one or more programs are executed by the apparatus, the apparatus acquires the track data and the identification information corresponding to the track data; generating a resident point sequence according to the track data; matching the resident point sequence with preset plot information to further generate a plot access sequence; performing feature extraction on the land parcel access sequence based on a preset target dimension to generate space-time features corresponding to each identification information; and generating a plot representation of each plot in the plot access sequence based on each spatio-temporal feature.

According to the technical scheme of the embodiment of the application, feature extraction is carried out on the land parcel access sequence based on the preset target dimension, and space-time features corresponding to each identification information are generated; and generating a plot representation of each plot in the plot access sequence based on each spatiotemporal feature. Based on the obtained plot image, fine-grained management of urban areas can be assisted, and support is provided for detecting the transportation of special substances, the abnormity of inter-area circulation, early warning of dangerous areas, accurate advertisement putting of areas, traffic flow analysis and urban planning, so that the efficiency and the accuracy of scoring and marking of multiple dimensions of the plot are improved, the manpower participation is reduced, and the data processing cost is reduced.

The above-described embodiments should not be construed as limiting the scope of the present application. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions can occur, depending on design requirements and other factors. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method of mapping a terrain, comprising:

generating a resident point sequence according to the track data;

performing feature extraction on the land parcel access sequence based on a preset target dimension to generate space-time features corresponding to the identification information;

and generating a plot representation of each plot in the plot access sequence based on each spatiotemporal feature.

2. The method of claim 1, wherein the generating the sequence of dwell points comprises:

and generating a resident point sequence according to the vehicle resident position and the resident duration.

3. The method of claim 1, wherein prior to the feature extraction of the sequence of block visits based on a preset target dimension, the method further comprises:

dividing the block access sequence based on natural day or residence time to obtain each sub-block access sequence; and

the feature extraction of the parcel access sequence based on a preset target dimension to generate a spatiotemporal feature corresponding to each identification information comprises:

and respectively extracting the characteristics of the access sequences of the sub-blocks based on the time dimension and the space dimension so as to generate the time characteristics and the space characteristics corresponding to the identification information.

4. The method of claim 1, wherein generating a parcel representation of each parcel in the parcel access sequence based on each of the spatiotemporal features comprises:

and determining corresponding identification information for each land parcel in the land parcel access sequence, and calling a pre-training feature extraction module in a deep learning network model to generate a land parcel portrait of each land parcel in the land parcel access sequence based on the spatio-temporal feature corresponding to the identification information corresponding to each land parcel in the land parcel access sequence.

5. The method of claim 4, wherein generating a parcel representation of each parcel in the sequence of parcel accesses comprises:

selecting a land parcel as an anchor point from each land parcel access sequence;

screening the land parcel access sequences passing through the anchor points, and further determining a target land parcel access sequence from the land parcel access sequences passing through the anchor points;

and acquiring the space-time characteristics corresponding to the target parcel access sequence, and generating a parcel portrait of a parcel corresponding to the anchor point based on the space-time characteristics corresponding to the target parcel access sequence, the position of the anchor point in the target parcel access sequence, the anchor point and the space-time characteristics of the parcel access sequence in which the anchor point is located.

6. The method of claim 5, wherein determining a target sequence of block accesses from the sequence of block accesses passing through the anchor point comprises:

determining the length of the block access sequence passing through the anchor point and the position of the anchor point;

and determining a target block access sequence from the block access sequences passing through the anchor point based on the length and the anchor point position.

7. The method of claim 4, wherein prior to said invoking a pre-trained feature extraction module in a deep learning network model, the method further comprises:

acquiring an initial neural network model;

acquiring a training sample set, wherein the training sample set comprises similar sample pairs based on spatial features, similar sample pairs based on temporal features, spatial anchor point pairs corresponding to the similar sample pairs based on the spatial features, time anchor point pairs corresponding to the similar sample pairs based on the temporal features, first distances of the space anchor point pairs, second distances of the time anchor point pairs and marked images of a land mass;

using the similar sample pair based on the spatial features as the input of a coding conversion layer of the initial neural network model, using the spatial anchor point pair corresponding to the similar sample pair based on the spatial features and the first distance as the expected output of the coding conversion layer of the initial neural network model, and performing self-supervision training on the initial neural network model;

using the time characteristic-based similar sample pair as an input of a coding conversion layer of the initial neural network model, using a time anchor point pair corresponding to the time characteristic-based similar sample pair and the second distance as an expected output of the coding conversion layer of the initial neural network model, and performing self-supervision training on the initial neural network model;

8. A block imaging apparatus, comprising:

a receiving unit configured to acquire track data and identification information corresponding to the track data;

the plot access sequence generating unit is configured to match the resident point sequence with preset plot information so as to generate a plot access sequence;

the space-time feature generation unit is configured to perform feature extraction on the land parcel access sequence based on a preset target dimension so as to generate space-time features corresponding to the identification information;

a plot representation generation unit configured to generate a plot representation of each plot in the plot access sequence based on each of the spatiotemporal features.

9. The apparatus of claim 8, wherein the dwell sequence generation unit is further configured to:

10. The apparatus of claim 8, further comprising a sequence partitioning unit configured to:

the spatio-temporal feature generation unit is further configured to:

11. An electronic device for data processing, comprising:

one or more processors;

a storage device for storing one or more programs,

when executed by the one or more processors, cause the one or more processors to implement the method of any one of claims 1-7.

12. A computer-readable medium, on which a computer program is stored, which, when being executed by a processor, carries out the method according to any one of claims 1-7.